Evolutionary many-objective optimization for mixed-model disassembly line balancing with multi-robotic workstations
Evolutionary many-objective optimization for mixed-model disassembly line balancing with multi-robotic workstations
复制标题
DOI：
10.1016/j.ejor.2018.12.035
复制
发表时间：
2019-07-01
影响因子：
6.4
通讯作者：
Duc Truong Pham
中科院分区：
文献类型：
--
作者：
Fang, Yilin;Liu, Quan;Duc Truong Pham
在再制造产业中，自动化拆卸成为最有希望获得经济效益的解决方案之一。机器人拆卸线平衡是实现工业规模自动化拆卸的关键问题。研究了多机器人工作站混流拆卸线平衡的进化多目标优化问题。在每个工作站中，多个熟练的机器人同时执行属于不同报废产品或机器人的不同任务。基于变换后的AND/OR图和并行拆卸，提出了一个数学规划模型，以最小化周期时间、总能耗、峰值工作站能耗和同时使用的机器人数量为目标。此外，问题的知识杠杆进化算法，包括编码/解码方案，初始化方法和特定问题的变异算子，开发来处理上述问题。基于本研究中生成的8个产品模型和63个问题实例进行了综合实验。特别是，我们提出的算法和5个代表性的进化算法从处理多目标问题的3类方法中选出的比较研究。然后，我们的问题的进化算法的设计方面的一些见解，从调查中获得。(C)2018 Elsevier B. V.版权所有。
In the remanufacturing industries, automated disassembly becomes one of the most promising solution in achieving economic benefit. Robotic disassembly line balancing is a key problem that enables automated disassembly to be implemented at industrial scale. This paper focuses on evolutionary many-objective optimization for mixed-model disassembly line balancing with multi-robotic workstations. In each workstation, multiple skilled robots perform different tasks belonging to the different end-of-life products or subassemblies simultaneously. Based on the transformed AND/OR graph and parallel disassembly, a mathematical programming model is proposed to minimize the cycle time, the total energy consumption, the peak workstation energy consumption, and the number of robots being used simultaneously. Furthermore, a problem knowledge-leveraging evolutionary algorithm, including encoding/decoding scheme, initialization approach and problem-specific variation operators, is developed to deal with the above problem. Comprehensive experiments are conducted based on 8 product models and 63 problem instances generated in this study. In particular, a comparative study of our proposed algorithm and 5 representative evolutionary algorithms selected from the 3 classes of approaches of dealing with many-objective problems are provided. Then some insights with respect to the design of evolutionary algorithms for our problem are gained from the investigation. (C) 2018 Elsevier B.V. All rights reserved.